The relative amplitude method is applied to body waves recorded at teleseismic distances from the October 30, 1994, mb 5.6 seismic disturbance, in South Africa, to determine the source depth and moment tensor. The results indicate that the source is shallow and that moment tensors with a positive volume change component are inconsistent with the observations. Observations of Rg in the frequency range 0.5--2.0 Hz at two regional distance stations also indicate that the source is shallow (less than about 3 km deep). Three-component broadband and long-period seismograms recorded at regional distances show strong Love wave excitation and a lobed Rayleigh radiation pattern, observations that are inconsistent with collapse mechanisms recently proposed for some large rock bursts (mine tremors). Modeling of teleseismic broadband P and regional distance surface waves shows that the source depth is shallow (2--3 km) and that the observed source radiation is consistent with that expected from a 45¿ normal dip-slip fault. The results of the modeling are confirmed by reports from the mining region that the disturbance was a mine tremor caused by slip on the Stuirmanspan Fault in the President Brand mine at a depth of between 2 and 3 km. The close match between observed and synthetic P and Rayleigh wave amplitudes suggests that the explosion-like mb:Ms of the mine tremor (mb 5.6, Ms 4.7) is attributable to the effect of the mechanism and the shallow depth of the source. The above results are at odds with those of Fan and Wallace <1995>, who interpreted this disturbance as an earthquake at a depth of 12 km, based on a broadband inversion of three-component waveforms from two regional distance stations. This study confirms the need to interpret both broadband waveforms (to match amplitude and polarity) and short-period waveforms (to confirm (relative) onset times in the model structure) if inversion of regional broadband waveforms is to provide a reliable estimate of the moment tensor and source depth.